While previous studies have primarily focused on the impact of artificial intelligence (AI) on labor markets, its effects on population migration and settlement behavior remain insufficiently explored. How does AI development reshape the long-term settlement intentions of migrants through labor market mechanisms? Does this effect exhibit significant heterogeneity across different skill levels which migrants have? To address these questions, this study draws on data from the China Migrants Dynamic Survey (CMDS) (2012-2018) and Baidu migration data (2019-2024) to systematically examine the impact of AI development on migrants' long-term settlement intentions.

The results indicate that AI development significantly increases migrants' long-term settlement intentions, this conclusion remains robust after various robustness checks and addressing endogeneity concerns. Mechanism analysis reveals that, at the micro level, AI development enhances migrants' long-term settlement intentions by raising their income levels and labor market participation, thereby improving economic returns and employment stability. At the city level, AI development fosters migrants' long-term settlement intentions by stimulating urban economic growth, optimizing public services provision, and enhancing urban amenities. Heterogeneity analysis further demonstrates that the positive effect of AI development is more pronounced among high-skilled and high-income migrants, as well as those engaged in non-routine cognitive tasks, whereas low-skilled, low- to middle-income migrants, and those performing routine, easily replaceable tasks benefit significantly lesser. Further analysis reveals that AI development also exerts a notable positive effect on population migration behavior.

This study contributes to the literature in three main ways. First, in terms of research content, it integrates AI development into the analytical framework of population migration by focusing settlement intentions, thereby deepening the understanding of the nexus between technological change and population dynamics. Second, regarding research design, unlike most existing studies that rely on industrial robot adoption as a proxy for AI, this paper extracts firms' business scope data from the National Enterprise Credit Information Publicity System. By leveraging Large Language Models (LLMs) for keyword filtering, it constructs a city-level indicator of AI enterprise density. This approach more accurately measures the practical application and industrialization of AI, overcoming the manufacturing bias of robot-based data. Third, from a research perspective, this paper moves beyond the conventional view of migrants as a homogeneous group. By focusing on skill structures, it reveals the heterogeneous settlement decisions under technological shocks, providing new empirical evidence for the evolution of demographic structures.

Theoretically, this study elucidates how AI influences migrants' settlement intentions through labor market channels and urban amenities, enriching the discourse on migration. Practically, it advocates for inclusive AI development policies and the establishment of universal, forward-looking lifelong learning and reskilling systems. Particular emphasis should be placed on supporting low- and middle-skilled and low-income groups, ensuring that the dividends of AI development are shared more broadly to promote the synergy between technological progress and high-quality population development.